(411g) Predicting Site-Specific Reactivity from Local Surface Properties

Rational design of nanostructured catalysts relies on a detailed understanding of the catalytic material, and in particular of the catalytic active site. Herein, we present an approach for predicting the catalytic behavior of a given structural motif based on the local variations of chemical properties. Taking inspiration from enzymatic processes, we examine how the proper weighing of properties such as the local electron affinity, ionization potential, and electrostatic potential at a site can describe its ability to catalyze a certain reaction step. We find that this is a promising path towards tailoring catalytic materials for a desired application, paving the way towards the inverse design of a catalyst based on an optimal active site.

The properties described above are evaluated on metal and transition metal oxide surfaces using density functional theory calculations, as exemplified in Figure 1. Figure 1 also demonstrates the general vision of our methods, where property information is taken from both the catalyst and the interacting adsorbate, and thereafter combined to describe the total interaction. The functional form for weighting the contributions of the properties varies depending on the character of the interacting compounds, going from linear relationships for simple cases to convoluted machine-learned non-linear functions for more complex interactions. Also the amount of information needed from neighboring sites varies with the interaction type. These aspects will be discussed in the presentation, addressing advantages and disadvantages of our methods in relation to other approaches.

To demonstrate our methods, we apply them to a few test cases including the adsorption and dissociation of small molecules such as CO, H₂, H₂O, CH₄, and NH₃ – important in numerous catalytic processes. We envisage that these methods will be of broad general use in surface science and catalysis, including both thermal and electrochemical applications.